func (fs *CamliFileSystem) fetchSchemaSuperset(br *blobref.BlobRef) (*schema.Superset, os.Error) {
blobStr := br.String()
if ss, ok := fs.blobToSchema.Get(blobStr); ok {
return ss.(*schema.Superset), nil
}
log.Printf("schema cache MISS on %q", blobStr)
rsc, _, err := fs.fetcher.Fetch(br)
if err != nil {
return nil, err
}
defer rsc.Close()
jd := json.NewDecoder(rsc)
ss := new(schema.Superset)
err = jd.Decode(ss)
if err != nil {
log.Printf("Error parsing %s as schema blob: %v", br, err)
return nil, os.EINVAL
}
if ss.Type == "" {
log.Printf("blob %s is JSON but lacks camliType", br)
return nil, os.EINVAL
}
ss.BlobRef = br
fs.blobToSchema.Add(blobStr, ss)
return ss, nil
}
func (mi *Indexer) populateFile(client *mysql.Client, blobRef *blobref.BlobRef, ss *schema.Superset) (err os.Error) {
if ss.Fragment {
return nil
}
seekFetcher, err := blobref.SeekerFromStreamingFetcher(mi.BlobSource)
if err != nil {
return err
}
sha1 := sha1.New()
fr := ss.NewFileReader(seekFetcher)
mime, reader := magic.MimeTypeFromReader(fr)
n, err := io.Copy(sha1, reader)
if err != nil {
// TODO: job scheduling system to retry this spaced
// out max n times. Right now our options are
// ignoring this error (forever) or returning the
// error and making the indexing try again (likely
// forever failing). Both options suck. For now just
// log and act like all's okay.
log.Printf("mysqlindex: error indexing file %s: %v", blobRef, err)
return nil
}
attrs := []string{}
if ss.UnixPermission != "" {
attrs = append(attrs, "perm")
}
if ss.UnixOwnerId != 0 || ss.UnixOwner != "" || ss.UnixGroupId != 0 || ss.UnixGroup != "" {
attrs = append(attrs, "owner")
}
if ss.UnixMtime != "" || ss.UnixCtime != "" || ss.UnixAtime != "" {
attrs = append(attrs, "time")
}
log.Printf("file %s blobref is %s, size %d", blobRef, blobref.FromHash("sha1", sha1), n)
err = execSQL(client,
"INSERT IGNORE INTO files (fileschemaref, bytesref, size, filename, mime, setattrs) VALUES (?, ?, ?, ?, ?, ?)",
blobRef.String(),
blobref.FromHash("sha1", sha1).String(),
n,
ss.FileNameString(),
mime,
strings.Join(attrs, ","))
return
}
func (mi *Indexer) populateFile(blobRef *blobref.BlobRef, ss *schema.Superset) (err os.Error) {
seekFetcher, err := blobref.SeekerFromStreamingFetcher(mi.BlobSource)
if err != nil {
return err
}
sha1 := sha1.New()
fr, err := ss.NewFileReader(seekFetcher)
if err != nil {
log.Printf("mysqlindex: error indexing file %s: %v", blobRef, err)
return nil
}
mime, reader := magic.MimeTypeFromReader(fr)
n, err := io.Copy(sha1, reader)
if err != nil {
// TODO: job scheduling system to retry this spaced
// out max n times. Right now our options are
// ignoring this error (forever) or returning the
// error and making the indexing try again (likely
// forever failing). Both options suck. For now just
// log and act like all's okay.
log.Printf("mysqlindex: error indexing file %s: %v", blobRef, err)
return nil
}
log.Printf("file %s blobref is %s, size %d", blobRef, blobref.FromHash("sha1", sha1), n)
err = mi.db.Execute(
"INSERT IGNORE INTO bytesfiles (schemaref, camlitype, wholedigest, size, filename, mime) VALUES (?, ?, ?, ?, ?, ?)",
blobRef.String(),
"file",
blobref.FromHash("sha1", sha1).String(),
n,
ss.FileNameString(),
mime,
)
return
}